CVX2 0.0.9

Tools of CV(Computer Vision)

Usage Sample ''''''''''''

.. code:: python

    import torch
    from torch import nn
    from cvx2 import WidthBlock
    from cvx2.wrapper import ImageClassModelWrapper

    model = nn.Sequential(
        WidthBlock(c1=1, c2=32),
        nn.MaxPool2d(kernel_size=2, stride=2),
        WidthBlock(c1=32, c2=64),
        nn.MaxPool2d(kernel_size=2, stride=2),
        nn.Flatten(),
        nn.Linear(in_features=64*49, out_features=1024),
        nn.Dropout(0.2),
        nn.SiLU(inplace=True),
        nn.Linear(in_features=1024, out_features=2),
    )

    img = torch.randn(1, 1, 28, 28)
    print(model(img).shape)

    data_dir
     |__train
     |   |__class1
     |   |   |__001.jpg
     |   |   |__002.jpg
     |   |__class2
     |      |__001.jpg
     |      |__002.jpg
     |__test
     |   |__class1
     |   |   |__001.jpg
     |   |   |__002.jpg
     |   |__class2
     |      |__001.jpg
     |      |__002.jpg
     |__val
         |__class1
         |   |__001.jpg
         |   |__002.jpg
         |__class2
            |__001.jpg
            |__002.jpg

    model_wrapper = ImageClassModelWrapper(model)
    model_wrapper.train(data='data_dir', imgsz=28)
    result = model_wrapper.predict('data_dir/test/class1/001.jpg', imgsz=28)